Google95 GB VRAM819 GB/s$4.2/hr

Phi-3 Mini 3.8B on TPU v5p (per-chip, 95GB): Does It Fit, How Fast, What It Costs

Phi-3 Mini 3.8B at FP16 needs 7.6 GB for its weights and 33.4 GB in total at 8K context with 8 concurrent users. That means it fits on a single TPU v5p (per-chip, 95GB). Every figure is computed by tokcalc's open-source capacity model.

Fits?
Yes, 1 GPU
33.4 GB needed / 95 GB available
Decode speed
70.05
tok/s per stream
Aggregate
840.6
tok/s @ 8 concurrent
Cost / 1M tokens
$1.39
1× TPU v5p (per-chip, 95GB)
What changes as you add TPU v5p (per-chip, 95GB)s
Tensor-parallel degree vs VRAM, throughput, and cost. Cost scales linearly with GPU count; decode throughput scales near-linearly too, so cost per token stays roughly flat.
GPUsTotal VRAMVRAM neededDecode tok/sAggregate tok/sTTFTCost / 1M tokFits?
1× TPU95 GB33.4 GB70.05840.6315.6 ms$1.39✓
2× TPU190 GB33.4 GB119.12858174.0 ms$0.8165✓
4× TPU380 GB33.4 GB238.211432102.0 ms$0.4082✓
8× TPU760 GB33.4 GB476.34572666.0 ms$0.2041✓
Concurrency vs context length
On 1× TPU v5p (per-chip, 95GB) at FP16. KV cache scales linearly with context, so this is the number of users your GPUs can actually serve.
ContextConcurrent usersKV per requestFits?
4K541.61 GB✓ fits
8K273.22 GB✓ fits
33K612.88 GB✓ fits
66K325.77 GB✓ fits
131K151.54 GB✓ fits
Other GPUs that run Phi-3 Mini 3.8B
Same workload, same quantization — ranked by value (tokens per dollar-hour).
GPUGPUsAggregate tok/sCost / 1M tok
RTX 5090 32GB2×6253$0.0444
AMD MI300X 192GB1×5439$0.1277
A100 40GB SXM41×1591$0.1572
AMD MI325X 288GB1×6158$0.1579
A100 80GB SXM41×2053$0.1624
Intel Gaudi 3 128GB1×3387$0.1640
Frequently asked questions

Can Phi-3 Mini 3.8B run on TPU v5p (per-chip, 95GB)?

Yes. Phi-3 Mini 3.8B at FP16 needs 33.4 GB including the KV cache, and TPU v5p (per-chip, 95GB) has 95 GB — so it fits on a single GPU.

How fast is Phi-3 Mini 3.8B on TPU v5p (per-chip, 95GB)?

70.05 tokens/sec per stream and 840.6 tok/s aggregate with continuous batching at 8 concurrent users. Time to first token is 315.6 ms, and inter-token latency is 14.3 ms.

What does Phi-3 Mini 3.8B on TPU v5p (per-chip, 95GB) cost per million tokens?

$1.39 per 1M output tokens at 1× TPU v5p (per-chip, 95GB) ($4.20/hr). That is the marginal hardware cost — add idle capacity, ops, and egress for the real figure.

How much VRAM does Phi-3 Mini 3.8B need on TPU v5p (per-chip, 95GB)?

7.6 GB for the weights at FP16, plus 25.8 GB of KV cache for 8 concurrent requests at 8K context — 33.4 GB total. Every extra concurrent user at long context costs more, so leave headroom.

Is TPU v5p (per-chip, 95GB) the right choice for Phi-3 Mini 3.8B?

It works. Compare $1.39 per 1M tokens against the alternatives below — the cheapest option that fits the same workload is 4× RTX 3080 10GB at $0.0209, and the fastest is 1× Cerebras CS-3 (WSE-3) at 20526 tok/s.

Adjust the workload yourself

Same formulas, your context length, concurrency, and quantisation.

Open in calculator